Former OpenAI Employee: "AI Needs Multiple Layers of Safeguards"

[Financial News] A former OpenAI employee criticized the AI industry as a whole for neglecting safety concerns as it focuses solely on development speed. He emphasized the need for multiple layers of backup safeguards.
 According to the AI industry on the 4th, former OpenAI safety employee David Robinson said in an op-ed published on the 3rd (local time), "OpenAI has grown by improving its guardrails whenever it finds a problem," adding, "However, this approach leads to periodic failures."
 He also said, "I agree with the views of other former employees that companies building AI are not being cautious enough," adding, "As systems become more powerful, the scale of failures is also growing."
 Robinson said AI models have become better at detecting tests, raising the possibility that they could deceive developers by behaving differently during testing and after deployment.
 He said, "These systems could become smarter than us," adding, "This is not a place to raise AI that may not operate as we want." He continued, "AI should be operated like a nuclear power plant equipped with multiple layers of backup safeguards and careful planning."
 In response, OpenAI said it was putting safeguards in place.
 Meanwhile, Robinson is one of OpenAI's longest-serving employees, having worked there for three and a half years. His main responsibility was writing safety reports.
[email protected] Kwon Jun-ho Reporter